基于串归约的网格工作流费用优化方法.pdfVIP

ISSNN 111777TP Journal of Computer Research and Development 5( 2) : 2 6~ 253, 2008 1, 3 1, 2 1,2 苑迎春 李小平 王 茜 1( 210096) 2 ( 210096) 3 ( 071001) (nd hd yyc@ 163com) Cost Optimization Heuristics for Grid Workflows Scheduling Based on Serial Reduction Yuan Yingchun1, 3, Li Xiaoping1,2, and Wang Qian1,2 1( School of Comp uter Science and Engineering , Southeast University , N anj ing 210096) 2 ( M inistry of Education K ey La oratory of Comp uter Network and I nf ormation I ntegration , Southeast University, N anj ing 210096) 3 ( Faculty of I nf ormation Science and Technology , A griculture University of H e ei, Baoding 071001) Abstract Workflow scheduling w hich guarantees the anticipated QoS ( quality of service) is a fundamental and complexity problem in computational grids In this paper, the scheduling of w orkflows represented by directed acyclic graph ( DAG) w ith the objective of cost minimization is considered In general, the opti mization problem is NPhard Tw o new heuristics, FSRD ( forw ard serial reduction w ith deadline) and BSRD ( backw ard serial reduction with deadline) are proposed According to the property of serial activi ties, a new concept called series reduction ( SR) is definedThe time window for each serial reduction group is recalculated based on leveling algorithms A dynamic programming method is presented for implementing the cost minimization of each serial reduction group By integrating the local optimization strategy with two leveling algorithms ( DBL and DTL) , BSRD and FSRD are investigated The tw o heuristics take into ac count comprehensive serial and parallel properties of DAG, and improve the cost optimization strategies adopted by leveling algorithms Experimental results show that BSRD and FSRD can con